Cook's distance is a measurement of the influence of the i^th data point on all the other data points.  In other words, it tells how much influence the i^th case has upon the model.  The formula to find Cook's distance, D_i, is

 where

  _j is the  predicted (fitted) value of the i^th observation;

_j(i) is the predicted value of the j^th observation using a new regression equation found by deleting the i^th case;  

 p is the number of parameters in the model;

and MSE is the Mean Square Error

Using the F distribution to compare with Cook's distance, the influence that the i^th data point has on the model can be found.  Values in the F distribution table can be used to express the percentage of influence the i^th data point has.  A percentage of 50% or more would indicate a large influence on the model.   The larger the error term implies that the D_i is also larger which means it has a greater influence on the model. 

Next, learn the procedure for calculating Cook's distance.

Regression Tutorial Menu    Dictionary

STATS @ MTSU